simon-5510-04-slides

2025-02-12

Topics to be covered

  • What you will learn
    • Experimental, quasi-experimental, observational studies
    • How to randomize
    • Advantages and disadvantages of randomization
    • Alternatives to randomization
    • Blinding
    • Intention to treat analysis
    • Baseline measures

Learning objectives

  • To define what a randomized study is and explain its advantages and disadvantages.

  • To describe how blinding, concealed allocation, and intention to treat analysis can improve the persuasiveness of a randomized trial.

Cameron and Pauling study of Vitamin C.

  • Linus Pauling
    • Two-time Nobel prize winner (Chemistry, Peace)
    • Fervent advocate of mega-doses of Vitamin C
  • Non-randomized study of cancer
    • All patients got Vitamin C
    • Ten controls per treated patient

Birth control pills and cervical cancer, 1

  • Three groups of women
    • Oral contraceptives
    • Other forms of birth control
    • No birth control

Birth control pills and cervical cancer, 2

  • Differences in these three groups
    • Age at first intercourse
    • Number of sexual partners
    • Use of tobacco, alcohol, other drugs
  • Most importantly, number of pap smears
    • Oral contraceptives require regular doctor visits

Experimental versus observational

  • Experimental: Active independent variable
  • Observational: Attribute independent variable
  • Who does the choosing?
    • Researcher = Experimental
    • Patients or doctors = Observational
  • Patients in a randomized trial are giving you a gift.

Randomized versus quasi-experimental

  • Quasi-experimental
    • Logistics prevent randomization
    • Very common in quality improvement

Break #1

  • What you have learned
    • Experimental, quasi-experimental, observational studies
  • What’s coming next
    • How to randomize

A fishy research story, 1

A fishy research story, 2

A fishy research story, 3

A fishy research story, 4

A fishy research story, 5

A fishy research story, 6

A fishy research story, 7

How to randomize

  • Use of a random device
    • Physical devices (flip of a coin, balls in an urn)
    • Computer generated random numbers
  • Avoid haphazard approaches
  • Avoid alternating assignment
  • Remember the law of large numbers

Simple randomization

  • Place the data in a systematic order
  • Add a column of random numbers

  • Sort the column of random numbers

Block randomization

  • Place the data in a systematic order
  • Add a column of random numbers

  • Sort by block and then the column of random numbers

Randomization in a cross-over design

  • Place the data in a systematic order
  • Add a column of random numbers

  • Sort by subject and then the column of random numbers

Random selection

  • Place the data in a systematic order
  • Add a column of random numbers

  • Sort by the column of random numbers
  • Select starting from the top

Break #2

  • What you have learned
    • How to randomize
  • What’s coming next
    • Advantages and disadvantages of randomization

Advantages of randomization

  • Insures covariate balance
    • Smoking during pregnancy and Down’s syndrome
  • Difficult or impossible to measure covariates
    • Severity of illness
    • Co-medications
    • Co-morbidities
    • Patient’s psychological state
  • Avoids selection bias

Disadvantages of randomization

  • Expensive
  • Artificial
    • Extra tests, extra attention
    • Explicit acknowledgement of uncertainty

Problems with randomized trials

  • Volunteer bias.
    • Willingness to endure painful procedures
  • Professional volunteers
  • Strong personal preferences
    • Birth control methods
    • Surgical versus non-surgical trials
    • Less invasive surgery

Break #3

  • What you have learned
    • Advantages and disadvantages of randomization
  • What’s coming next
    • Alternatives to randomization

When can’t you randomize, 1

  • Unethical
    • Lack of equipoise
    • Harmful exposures

When can’t you randomize, 2

  • Impractical
    • Too many patients
    • Too much time
    • Strong patient preferences

When can’t you randomize, 3

  • Impossible
    • Attribute variables
    • Retrospective studies

Some alternatives

  • Wait list control groups
  • Add on trials
  • Intensive versus standard advice
  • Randomly assign from a different end
  • Early stopping

Break #4

  • What you have learned
    • Alternatives to randomization
  • What’s coming next
    • Blinding

Blinding/partial blinding. Who knew what when?

  • Hiding information (not deception)
  • Not always possible (bilateral orchiectomy)

Types of blinding

  • Double blind
    • Physician and patient blinded
  • Single blind
    • Patient only
  • Partial blind
    • Evaluators blinded

Hawthorne effect

  • Series of studies at a GE factory.
    • Any change, no matter what, improved productivity
    • Positive response to attention.

Ascertainment bias

  • The tendency to self deception.
    • “Linus Pauling actively promoted the use of massive doses of vitamin C during the last few decades of his life. He believed it could cure just about anything from the common cold to cancer. During one interview he explained that after he and his family started taking Vitamin C supplements, they never had colds. The interviewer was a bit surprised probed a bit further ‘No colds? Ever?’ Linus Pauling responded, ‘Oh just an occasional sniffle.’”

Confusion about the placebo effect, 1

  • Natural course of a disease
    • “If a doctor treats your cold, it will go away in fourteen days. If you leave it alone, it will go away in two weeks.” Gloria Silverstein.
    • “The art of medicine consists in amusing the patient while nature affects the cure.” Voltaire

Confusion about the placebo effect, 2

  • Regression to the mean
    • You’re never as good as you think you are on your good days and you’re never as bad as you think you are on your bad days.
  • Hróbjartsson and Gøtzsche study of placebo effect

Subversion of the randomization process, 1

  • Physician subversion
    • Waiting until the right number pops up
    • Biased implementation of exclusion criteria
    • Not a problem in single investigator trials
  • Hiding the randomization list.
    • Sealed envelopes
    • 800 number

Subversion of the randomization process, 2

  • Patient subversion of the randomization process
    • Early anti-retroviral trials for AIDS

Break #5

  • What you have learned
    • Blinding
  • What’s coming next
    • Intention to treat analysis

Intention to treat analysis

  • Controversial
  • Analyze in groups originally randomized to
    • Maintains benefit of randomization
    • Measures practical impact of treatment
  • Alternative: per protocol analysis

Break #6

  • What you have learned
    • Intention to treat analysis
  • What’s coming next
    • Baseline measures

Baseline measurements

  • Not an absolute requirement
  • Reduces variance
  • Allows quality checks
    • Subversion of randomization
    • Differential dropout
  • Several competing analysis strategies

Summary

  • What you have learned
    • Experimental, quasi-experimental, observational studies
    • How to randomize
    • Advantages and disadvantages of randomization
    • Alternatives to randomization
    • Blinding
    • Intention to treat analysis
    • Baseline measures
  • Next week: Quasi-experimental designs